ChatGPT and Google AI are Transforming the Internet - Matthew Prince | Radical with Amol Rajan

I'm here for a few days. Yeah. Yeah.

Yeah. So, the internet's going through a massive transition right now, and I think that transition is inevitable because at the end of the day, it's better for users. And that transition is that we're moving from an internet where as you tried to discover things on it, you used search engines. And we're all familiar with Google or Bing or the various search engines that are out there. And in search engines, you would type a query in and then it would give you back essentially a treasure map of 10 links. And the search wasn't over when you got the map. You then as a user had to click on those links and search around to find whatever answer it is. So if you wanted to see, you know, when was Cloudflare founded, you type that in, I give you back 10 links. You go click on it, you'd read through various articles, and then you'd find the answer to that. And in that process of like engaging and searching on that treasure map, you generated traffic and then content creators like yourself and others could monetize that through advertisements, through subscriptions, or just again as you said, a lot of the content online is done just because you want to know that what you're putting out in the world is actually being consumed. And so, you know, my wife and I own a small newspaper. Our writers don't get paid based on how many people read their articles, but they're all obsessed with, you know, how many page views am I getting? How many people are coming and reading my thing? Because that's part of what's important about being a content creator.

Yeah. And I think that that's just part of the experience that we've had online. And that deal that you would put content out there and in exchange you could either make money or you could get famous from having done that. That deal is what has generated the internet that we have today. But that deal is changing and you can actually see it by just watching any sci-fi movie. Some sci-fi movie you imagine has a helpful robot and you say I'd like a recipe for cookies. The robot doesn't come back and say here are 10 links and maybe you can find it somewhere. The robot comes back and says here's a recipe for cookies. And that once upon a time was science fiction. But if you look at the AI companies today, you look at something like ChatGPT, you look at Anthropic, you look even at what Google is becoming where they're putting AI overviews at the top of the page. Instead of it giving you a treasure map, instead of it being a search engine, this is becoming an answer engine. And the difference of that is that if you want to say, okay, what did Amal talk about on his podcast today? The answer engine will tell you, but there's no human who has had to go and actually listen to your podcast. I mean, there's a commercial consequence of that, which is that the lack of traffic means that all those people can't monetize the reach. There's also just a sort of almost a neurological consequence which is that you're not having to do any thinking. I mean the AI overview is kind of giving you an answer and previously if you wanted to work out when Cloudflare was set up you might go to the Cloudflare website you might go to Wikipedia you might go to two others you might verify you might do some thinking and of course if most humans or most of human history would suggest if humans can opt out of doing some thinking they're quite keen to do that. But how did you come to realize that this was a growing phenomenon and actually the complaints people were maybe making to you which is hey Matthew you know I'm using your stuff but you know people aren't coming to me as much as they used to. Suddenly you realize actually they're telling the truth.

Yeah. I think that so first of all I think you're right that this is exactly it is inevitable that this is going to be the case because again whether you're thinking or not you're getting the answer faster and you're getting what you want a little bit faster and that for 95% of users 95% of the time is a better experience. It's convenient and so people are going this is what the future is going to look like. I don't think we can hold back progress. There are going to be answer engines and they're going to replace search engines and that's what the future is going to look like. What we started to hear because we have so many clients, is at first media companies coming to us and just saying we're getting killed. And our business primarily is to protect against cyber attacks, you know, the Chinese or the Russians or the North Koreans or the Iranians, you know, hacking folks. And so when they were like, we have this new threat and it's AI companies, my first reaction was, oh god, it's media companies that, you know, roll my eyes. They're just complaining about new technology again. They're such Luddites.

Yeah. And so, you know, I think that that was my first reaction. And company after company after company said, 'Just go pull the data.' And so we did. And people might not totally understand how search engines work, but the way that they work is essentially they have a little robot or what might be called a crawler or a bot that's going around and visiting everything on the internet. And so Google every day is making, you know, billions of requests to access all of the content of the internet. And it gets a copy of everything that has been created and then it uses that to build an index which is what the search engine is and then when you search it returns the results that you see. And so the deal has been that you let Google copy your content in exchange for appearing in their treasure map that they have the visibility.

And so we can see for the last 10 years what the ratio of how much Google crawls that content, Google takes it, versus how many page views, how many visitors they actually send you. And the ratio 10 years ago was they crawled two pages for every one visitor that they sent you. So they crawled two, they sent you one visitor. Over that period of time that has gotten much much worse. So today it's up to almost 20 to one. They crawl 20 pages for every one visitor. What's changed? AI overviews because that then says I don't have to click on anything. I don't have to go anywhere. I get the answer right there. And that's the good news for publishers. If you look at folks like OpenAI with ChatGPT, they're at 1,500 crawls for every one visitor that they send you. If you look at something like Anthropic, it's 40,000 crawls for every one visitor they send you. And just from a fairness perspective, like that's a cost to the BBC. That's a cost to content creators. They have to pay for all of that crawled content. And yet they're not getting anything in exchange. They're not getting the traffic. And the world over time is going to move to these answer engines. And as it does, it means that the deal that content creators have with them, it has to change.

Sure. And again, I think we can't dispute that these products are good, that for a lot of consumers, they're providing real value. But the problem is you need to eat, right? Something needs, there needs to be some business model where content creators are compensated for what they are delivering. And if it's not going to be from traffic, it needs to be from something else. And I think there, you know, there are sort of three possible futures that we could have. The bleakest is you starve to death and die, right? Because nobody is actually paying you to create content anymore because it's just getting all sucked up into the answer engines. I think that's pretty unlikely because the answer engines need the content. It is incredibly valuable for them to have that content. It's the fuel that actually makes these things run. And so in the end they've got an incentive to make sure that we can feed ourselves.

I think that the sort of Black Mirror potential future, which is really scary because I think it's pretty likely, is that instead of there being, you know, hundreds of thousands of independent content creators, that we don't go back to a media time of, you know, 30 years ago, we actually go back to a media time of the 1400s, the Medici, where there are five powerful families and if you want to, you know, paint the Sistine Chapel, you've got to be associated with one of those families, but instead of them being families, they'll be AI companies. And again, there will be a liberal one, there will be a conservative one, there will be a Chinese one. And you can imagine that they will be the ones who decide what content gets created. They will essentially be the patrons of content creation. And it's not crazy to think about like how can you imagine Sam Altman from OpenAI deciding to stand up his own version of the Associated Press. A lot of out of work journalists these days would cost not that much relative to what they're spending on GPUs or AI use, the general processing units which is what powers that silicon that is there where they're literally spending 10 billion dollars a year. You could have a lot of newspapers for 10 billion dollars a year. And that I think is actually a really potential outcome that we have. But think about how regressive that is. If the internet was all about democratizing information, about spreading it, making accessible to anyone, making it so that everyone could go in any direction. If all of a sudden content gets siloed behind a small handful of different AI companies and your view of the world is completely dictated by which one you are subscribed to, that I think is actually an even more polarized and even more dangerous world than the one that we're living in.

Well, so there's the you don't eat outlook, that's the bleakest. There's the Black Mirror outlook which is we go back to sort of the five, you know, powerful AI companies that control all that. I think that what we're trying to play for is something else which is okay, traffic isn't going to be the currency anymore. It has to be something else. And so how do we create incentives where the AI companies say, you know, we need to pay content creators for the content that they're creating in order for our systems to have access to it. And there needs to be some value exchange where if you are creating content that you're not going to get as much ad revenue, you're not going to get as much subscription revenue, but let's create a new source of revenue, which is some fraction of every person who's subscribing to an AI company. Some fraction of that goes back to the content creators, which is the fuel that makes those AI companies work.

Yeah. So the argument that the News Corps of the world made against the Facebooks and Googles actually always kind of rang hollow to me because they were at least sending traffic back. They were generating that. Now was it as much traffic as they would have gotten before or were they at some level disintermediating the relationship with the reader or the viewer maybe, but the threat that the AI companies pose is so much larger. Again, 40,000 copies of your stuff for every one visitor they send you back to two to one.

So that argument that was before, I never didn't have much sympathy to, but this one I think is a real existential threat to the publishers first but actually to a lot of other businesses as well. What people don't appreciate is that AI is really a platform change. So if we start with the web and then after the web we had social and then after social we had mobile, each of those were platform changes. They were changing the interface through which you consumed information. So back in once upon a time it was on your laptop and then it was through a social media site and then it was through your mobile phone and apps on your phone and each of those was a platform change that was possible for people to see. What I don't think people appreciate is AI is a similar platform change where in the future the way that you're going to get information is you're going to ask your helpful chatbot and we're already seeing people move to that. What's different though about this platform is that this is a platform with an opinion for the first time. And what that's going to do is it's going to sit between you and yes, media companies first, but what happens to anyone who, what happens to a Booking.com? Like what value do they have anymore? Anyone who's aggregating information together? What happens to a brand? Like what do brands mean in a world where you tell your agent, 'Hey, go buy me eggs or go buy me milk or go buy me something like that.' There's no relationship between you as the human and what were effectively shortcuts for quality which is what brands are.

I mean, I think you're already being able to do things like how do I pull together a family vacation? You know, where should we stay, where should we eat, what activities should we do? And right now, for the most part, that's just pulling together here's it. And then you have to go out and buy it. But it is an inevitable step that the agent will then do that buying for you. And again, that's good. That's an amazing experience, but it's also going to change fundamentally business models. And so what I think is encouraging about this is traffic, which is the business model that the internet has been built on so far, has always been a really poor proxy for quality. Like there's a lot of things someone's like in a car wreck on the side of the road, like we all stop and stare at it. That's not good. That's not us at our best. Whereas what we're seeing from AI is that AI has actually for the first time in human history created effectively a mathematical model of all of human knowledge. And at the same time it has also shown us where there are gaps and holes in human knowledge. I picture like a giant block of Swiss cheese, there's a lot of cheese but there's a lot of holes in the cheese. And going forward if we can, if the business model of content creation, business model of many things is about to significantly change as the platform of AI shifts to how we consume information, how we shop, how we do everything online. As we do that, there's an opportunity for us to fix some of the problems of the last business model where again traffic and a lot of how media organizations have made money today is writing headlines that make us angry, that make us have to click on them, that divide us apart. Whereas if you imagine a future where the business model is how do we find those holes in the cheese and we actually fill them and the people who create that content get rewarded, that's actually a really optimistic future.

So the best thing I've kind of come up with is an analogy to when other parts of media were really disrupted. I like the Swiss cheese analogy. Swiss cheese is good. That's good. But look at the music industry. So 22 years ago before Steve Jobs steps on stage and launches iTunes 99 cents a song, the entire music industry, the market cap, how much it was worth, was about 8 billion US, which is a lot of money but I mean that's the Beatles and the Rolling Stones and I mean that's everything, 8 billion. And then at the time piracy was rampant, people were just sort of saying music is getting devalued. You had Napster, you had all these things. And then Apple launches iTunes 99 cents a song. Says, 'We're going to compete with the pirates. We're going to come up with a new business model.' Turns out that's not the business model that ends up winning. But you kind of have to go through that stage in order to get to Spotify. Spotify is, you know, closer to, you know, 10 pounds a month and all you can eat access to things. And Spotify alone last year sent 10 billion dollars US dollars to music creators around the world. And that's just Spotify. There's also Apple Music and Tidal and YouTube and TikTok. So the music industry is worth way more because technology has invented this new business model for it than it was before. And there are different winners and losers. And you can, there are people who are angry at Spotify because they say, 'Well, we should be being paid more or less,' but you can't dispute the fact that there is more money going into creating music today than there was just 20 years ago.

No one has compensated creators at scale like Daniel. It's just incredible. And I was talking to him and he told me this story that has just stuck with me, which is he said, you know, if you're on Spotify and you search for Taylor Swift, Shake It Off, like they know they're going to return a result that is what you're looking for. On the other hand, if you search for something like I want a song to a disco beat about how much fun it is to dance with your cat, they don't have much, but they know they don't have that much. They know that they've effectively found one of the holes in the cheese. And so what they do is they actually take that query and they publish it back to music creators. And there are several, like many, music creators that are making millions of dollars a year. He told me about one guy who's making 40 million euros a year just writing songs for unfulfilled Spotify queries. And I think that's beautiful because when you search for a song, what you're searching for is an emotion. And if you can't find it, then there's sort of a hole that's left. And these people are actually writing content to fill those holes. And I think that provides the beginnings of the ingredients for how we can remake the media industry in an age of AI to say what's really valuable is filling in the holes in the cheese. And that's I think the beginning of how we figure this out. And what I'm encouraged by is you talk to the founders of, you talk to the founders of OpenAI, Google, Microsoft, websites that you are a critical part of the infrastructure of.

Yeah. Oh, we see about 6 billion people pass through our network every month, which is essentially the entire online population.

What do I, what Cloudflare do? We make the internet faster and protected from bad guys. And so when we're doing our job right, your internet just seems a little bit faster. You're less likely to get scammed online. And we protect, yeah, most of our customers are either content creators or companies that are putting something making it accessible online and then their customers are the ones that flow through our network to be able to access whatever they're on.

The first meeting I'd ever been in that the term death squad had been used. Unfortunately, not the last.

We're about getting the internet to people who are going to otherwise be...